netdev - Re: [PATCH v2 net-next] tcp: set SOCK

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:	Fri, 15 May 2015 07:50:23 -0700
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	Jason Baron <jbaron@...mai.com>
Cc:	David Miller <davem@...emloft.net>, netdev@...r.kernel.org
Subject: Re: [PATCH v2 net-next] tcp: set SOCK_NOSPACE under memory pressure

On Fri, 2015-05-15 at 10:20 -0400, Jason Baron wrote:
> On 05/14/2015 11:07 PM, Eric Dumazet wrote:
> > On Sat, 2015-05-09 at 17:38 -0400, David Miller wrote:
> >> Applied, thanks a lot for adding so much information to the commit
> >> message.
> >> --
> > When doing more tests with this stuff, I found we could still have
> > hangs, because a TCP socket can end up having no skb in its write queue.
> >
> > We need to allow one skb (this can be 2KB) to keep the logic being ACK
> > driven.
> >
> 
> Indeed. I alluded to that in the commit message.
> 
> The 2KB minimum wouldn't help in the case where we really got
> -ENOMEM from the core allocator, unless I guess we reserved
> the 2KB permanently. Or we could just issue the wakeup in the
> case where we return -EAGAIN and the write queue is empty.
> That is a very simple additional change here, I think.

I cooked a patch series, will send it very soon.

The ENOMEM in core allocator is very unlikely, as we use GFP_KERNEL.

Normally, memcg and/or tcp_mem[] should prevent this from happening.



--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html